Doctoral Dissertation Research in Economics: Reference-Dependent Effort Provision under Heterogeneous Gain Loss Attitudes

理解个体在面对不确定性时如何做出决策是经济研究中的一个核心问题。虽然早期的经济理论假设完全理性，但研究人员通过引入参考依赖决策的思想，在这些经典模型的基础上进一步发展。这些模型根植于个人将其结果与某个参考点进行比较的直觉，已被用于许多情境，从职业高尔夫球手的推杆行为（Pope和Schweitzer， 2011）到失业工人的求职行为（DellaVigna等人，2017）。然而，原始的依赖于参考的模型有一个主要的缺点：由于没有指定一个固定的参考点，它们留下了相当大的自由度。理论的进步试图关闭这个问题，最终在基于期望的参考依赖模型（值得注意的是，Koszegi和Rabin， 2006 - KR）。通过将参考点内化为理性预期，该模型产生了新的、可测试的含义，获得了混合的实验支持。然而，这些先前的研究可能被普遍的损失厌恶假设所混淆，在这种假设中，低于参考点的结果被认为比高于参考点的结果感觉更糟。与这种假设相反，经验和实验证据表明，大约30%的人是“爱收获”的。重要的是，作者之前的工作表明，少数喜欢收益的参与者可能会扭曲模型的总体预测（Goette et al, 2019）。在这个项目中，研究小组将提供额外的证据，证明这些得失态度的异质性是如何混淆KR测试的，并设计一个实验范式来克服努力提供背景下的这些混淆。在考虑一些关键的政策问题时，更好地了解得失偏好的分布尤为重要，因为这些依赖于参考的模型已被证明在失业保险和健康保险决策等环境中非常宝贵。为了实现既定目标，本项目将着重于更新现有的实验范式，以解释得失态度的异质性。因此，提出了一个两阶段的实验，其中参与者的得失态度的测量从第一阶段的选择中恢复，并在第二阶段的假设进行测试。具体来说，参与者首先会被问到这样的问题：“在工资X的情况下，你愿意做多少项工作？”工资要么是确定性的（例如，每项任务20美分），要么是随机的（例如，有50%的几率，每项任务的工资是10美分，有50%的几率是30美分）。这种变化将使作者能够建立一个结构模型，并恢复得失态度的关键行为参数以及努力成本函数。在第二阶段，参与者被问及一个略有不同的问题，改编自Abeler等人（2011）：“如果你的工资是50%的机会每项任务25美分，50%的机会Y美元，无论任务的数量，你愿意工作多少？”关键的待遇将是随机变化的Y从小（5美元）到大（20美元）。KR预测，随着Y的增加，厌恶损失的代理人会更努力地工作，而喜欢收益的代理人会更努力地工作。通过使用先前测量的得失态度，该项目能够直接测试具有不同偏好的个体是否与理论预测的不同反应。由此产生的证据将有助于澄清先前范式的混合结果，并提供证据，证明在控制了得失态度的混杂异质性后，KR是否可以预测行为。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Understanding how individuals make decisions in the face of uncertainty is a core issue in economic research. While earlier economic theory postulated full rationality, researchers have advanced upon these classical models by introducing the idea of reference-dependent decision making. These models, rooted in the intuition that individuals compare their outcomes to some reference point, have been used in a number of contexts, ranging from the putting behavior of professional golfers (Pope and Schweitzer, 2011) to the job search behavior of unemployed workers (DellaVigna et al, 2017). The original reference-dependent models suffered from one major drawback, however: in failing to specify a fixed reference point, they left open a substantial degree of freedom. Theoretical advancements sought to shut this downfix this problem, culminating in models of expectations-based reference dependence (notably, Koszegi and Rabin, 2006 – KR). By endogenizing the reference point as rational expectations, this model generated new and, testable implications, garnering mixed experimental support. However, these prior studies are potentially confounded by an assumption of universal loss aversion, in which outcomes falling below the reference point are assumed to feel worse than those above the reference point. Contrary to this assumption, empirical and experimental evidence suggests that roughly 30% of individuals are instead “gain -loving”.” Importantly, prior work by the authors suggests that a small minority of gain -loving participants can skew aggregate predictions of the model (Goette et al, 2019). In this project, the research team will provide additional evidence of how heterogeneity in these gain-loss attitudes confounds tests of KR, and design an experimental paradigm to overcome these confounds in the context of effort provision. A better understanding of the distribution of gain-loss preferences is particularly important when considering a number of key policy questions, as these reference-dependent models have proven to be invaluable in settings such as unemployment insurance and health insurance decisions.In order to accomplish the stated goals, this project will focus on updating an existing experimental paradigm so as to account for the confound ofthat is heterogeneity in gain-loss attitudes. A two-stage experiment is thus proposed, wherein a measure of the participant’s gain-loss attitude is recovered from first stage choices, and the hypothesis is tested in the second. Specifically, participants will first be asked questions of the form: “How many tasks are you willing to work at wage X?”. The wages will either be deterministic (e.g. 20 cents per task) or stochastic (e.g. with 50% chance, the wage will be 10 cents per task and with 50% chance it will be 30 cents per task). This variation will allow the authors to build a structural model and recover the key behavioral parameter of gain-loss attitudes alongside a cost of effort function. Participants are asked a slightly different question in the second stage, adapted from Abeler et al (2011): “How much are you willing to work if your wage is a 50% chance of 25 cents per task, and a 50% chance of Y dollars regardless of the number of tasks?”. The key treatment will be randomly varying Y from a small ($5) to a large amount ($20). KR predicts that loss averse agents should work harder as Y increases, while gain loving agents should work less hard. By using the previously measured gain-loss attitude, this project is able to directly test whether individuals with different preferences respond differently as predicted by the theory. The resulting evidence will help clarify the mixed results from the prior paradigm, as well as providing evidence of whether KR is predictive of behavior after controlling for the confounding heterogeneity in gain-loss attitudes.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.